Add potential relevant interactions using step()
Description
![[Experimental]](data:image/svg+xml;base64,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)
Add potential relevant interactions to a model using stats::step(). The
function extract the formula of the model, identifies all potential
interactions and pass them as the upper component of the scope argument
to stats::step(). The current model formula is passed as the lower
component of scope.
Usage
add_interactions_by_step(model, ...) ## Default S3 method: add_interactions_by_step(model, ...)add_interactions_by_step(model, ...) ## Default S3 method: add_interactions_by_step(model, ...)
Arguments
model |
A model object. |
... |
Additional parameters passed to |
Value
The stepwise-selected model.
Examples
mod <- glm(as.factor(Survived) ~ ., data = titanic, family = binomial()) mod |> add_interactions_by_step()mod <- glm(as.factor(Survived) ~ ., data = titanic, family = binomial()) mod |> add_interactions_by_step()
Combine answers of a multiple answers question
Description
Considering a multiple answers question coded as several binary variables (one per item), create a new variable (list column or character) combining all positive answers. If defined, use variable labels (see examples).
Usage
combine_answers(data, answers, into, value = NULL, sep = NULL)combine_answers(data, answers, into, value = NULL, sep = NULL)
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
answers |
< |
into |
Names of new variables to create as character vector. |
value |
Value indicating a positive answer. By default, will use the maximum observed value and will display a message. |
sep |
An optional character string to separate the results and return a
character. If |
Note
If NA is observed for at least one item, return NA.
Examples
d <- dplyr::tibble( q1a = sample(c("y", "n"), size = 200, replace = TRUE), q1b = sample(c("y", "n", "n", NA), size = 200, replace = TRUE), q1c = sample(c("y", "y", "n"), size = 200, replace = TRUE), q1d = sample("n", size = 200, replace = TRUE) ) d |> combine_answers(q1a:q1d, into = "combined") d |> combine_answers(q1a:q1d, into = "combined", sep = ", ", value = "y") d |> combine_answers(q1a:q1d, into = "combined", sep = " | ", value = "n") # works with survey objects d |> srvyr::as_survey() |> combine_answers(q1a:q1d, into = "combined")d <- dplyr::tibble( q1a = sample(c("y", "n"), size = 200, replace = TRUE), q1b = sample(c("y", "n", "n", NA), size = 200, replace = TRUE), q1c = sample(c("y", "y", "n"), size = 200, replace = TRUE), q1d = sample("n", size = 200, replace = TRUE) ) d |> combine_answers(q1a:q1d, into = "combined") d |> combine_answers(q1a:q1d, into = "combined", sep = ", ", value = "y") d |> combine_answers(q1a:q1d, into = "combined", sep = " | ", value = "n") # works with survey objects d |> srvyr::as_survey() |> combine_answers(q1a:q1d, into = "combined")
Variable contribution and Dominance analysis
Description
Use an analysis of variance or an analysis of deviance to estimate the
contribution of each variable in reducing variance or deviance (sort of
R2 decomposition). This contribution could be explained as a
proportion of the total deviance (total contribution), of the residual
deviance (partial contribution) or of the explained deviance
(relative contribution), see details for more information.
contributions() computes the different contributions.
tbl_contributions() displays the results as a formatted gt
table, taking into account variable labels. Alternatively, tbl_domimance()
could be used to perform a dominance analysis with
dominanceanalysis::dominanceAnalysis() and display the results as a
formatted gt table.
Usage
contributions(mod, type = c("II", "III", "I", 1, 2, 3, "drop1", "add1"), ...) tbl_contributions( mod, type = c("II", "III", "I", 1, 2, 3, "drop1", "add1"), ..., show = c("Total", "Partial", "Relative"), decimals = 1, notes = TRUE ) total_deviance(mod) tbl_dominance(mod, indice = NULL, decimals = 1, totals = TRUE, notes = TRUE)contributions(mod, type = c("II", "III", "I", 1, 2, 3, "drop1", "add1"), ...) tbl_contributions( mod, type = c("II", "III", "I", 1, 2, 3, "drop1", "add1"), ..., show = c("Total", "Partial", "Relative"), decimals = 1, notes = TRUE ) total_deviance(mod) tbl_dominance(mod, indice = NULL, decimals = 1, totals = TRUE, notes = TRUE)
Arguments
mod |
a statistical model |
type |
type of Anova, Roman numerals being equivalent to the
corresponding Arabic numerals: |
... |
additional parameters passed to |
show |
list of contributions to display |
decimals |
number of decimals for deviance and contributions |
notes |
should table notes be added? |
indice |
fit indice name, see |
totals |
add a summary row with totals |
Details
In linear regression, the squared multiple correlation, R2 is used
to assess goodness of fit as it represents the proportion of variance in the
criterion that is explained by the predictors. It could be expressed as
1 - RSS / TSS where TSS represents the total sum of squares (the overall
observed variance) and RSS represents the residual sum of squares (i.e.
the variance not explained by the full model).
For generalized linear models (GLM), model fitting does not rely on ordinary
least squares (OLS). It is achieved by maximum likelihood. Therefore, sum
of squares is not available and R2 cannot be computed.
An alternative goodness of fit measure used in such case is pseudo
R2. The' McFadden pseudo R2 (sometimes called
likelihood ratio index) could be expressed using deviance rather than
likelihood. In such case, it is equal to 1 - Dr / Dt where Dt represents
the full deviance (i.e. the deviance of the null model, a model with no
predictor) and Dr the residual deviance (i.e. the deviance of
the full model). The McFadden pseudo R2 corresponds the
proportion of deviance reduced by the model. Deviance is a generalization of
the idea of using the sum of squares of residuals in the cases where
model-fitting is achieved by maximum likelihood.
In R, the residual deviance (Dr) of a GLM could be obtain with
stats::deviance(). For a linear model, stats::deviance() reports RSS.
The function total_deviance() could be used to get the total deviance
(Dt), i.e. the deviance of the null model. For a linear model, it provides
TSS.
Analysis of variance (Anova) is common to identify if the different
predictors have a significant effect on a model. Anova approaches have been
extended to also covers analysis of deviance. In the context of Anova-like
tests, it is common to report effect sizes indicators representing the amount
of variance or deviance explained by each variable included in the model.
Such approach is also known as R2 decomposition.
These indicators are based on Dpred (the delta of deviance) or SSpred
(the delta of the sum of squares, i.e. the delta of variance) reduced by the
inclusion of a specific predictor in the model.
A first measure, known as Eta-squared (η2) in the context of
linear models, expresses this delta of sum of squares (variance) as a
proportion of the total variance (SSpred / TSS). In the context of GLMs,
the delta of deviance could similarly expressed as a proportion of the
total deviance (Dpred / Dt), also known as the
semi-partial pseudo-R2. This indicator represents
the total contribution of a predictor in the reduction of deviance.
An alternative, known as partial Eta-squared
(ηp2), in the context of linear models, could
be expressed as SSpred / (RSS + SSpred), the variance uniquely explained
by the effect expressed as the proportion of variance not explained by the
other effects. Here the variance explained by the other effects in the model
is completely partialed out. Similarly, for GLMs,
partial pseudo-R2 is expressed as Dpred / (Dr + Dpred),
where Dr represents the residual deviance of the full model.
This partial contribution is the proportion of partial
variance/deviance uniquely explained by the associated effect.
Finally, it is possible to express a relative contribution as the
proportion of the variance/deviance explained by the model, that could be
be expressed as TSS - RSS or Dt - Dr. Therefore, relative contribution is
equal to SSpred / (TSS - RSS) or Dpred / (Dt - Dr) and represents the
relative reduction of variance/deviance of the predictor compared to the
total reduction of the variance / deviance by the full model.
It is crucial to understand the different types of Anova.
In a type-I Anova,
as performed by stats::anova(), the different predictors are included
sequentially and in-order into the model. Such analysis is therefore
order-dependant. The effect of a predictor is computed once taken into
account the previous predictors (but not the other one introduced later).
The first factor is tested without adjustment. The second factor is tested
after removing the effect of the first. The third is tested after removing
the effect of the first and the second. The advantage of such approach is
that the sum of relative contributions is equal to 100%. However, the
contribution of each variable changes if the variables are entered in a
different order.
Type II-Anova (default of car::Anova()) tests each main effect adjusted
for all other effects of the same order or lower, but not for interactions
involving that factor. Each main effect is tested as if it were the last
main effect entered. Type II are order-independent for main effects and are
generally preferred when there is no interaction. They have higher power
than Type III for testing main effects because they do not adjust for the
interaction term.
Type III-Anova (also done with car::Anova()) tests each effect adjusted
for all other effects in the model, including higher-order interactions.
Each effect is tested as if it were the last one entered into a model
containing all other effects. Type III are order-independent.
They require a specific contrast coding (typically sum-to-zero or Helmert)
to be interpretable.
car::Anova() calculates type-II or type-III Anova tables
indicating for each predictor the variance or deviance explained
by adding this predictor in the model compared to a model without this
predictor (but keeping all other predictors in the model).
contributions() also includes two alternatives: "drop1" and "add1".
type = "drop1" uses stats::drop1() which is equivalent, in the absence,
of interaction terms to a type-II Anova. For some models (such as
survey::svyglm() models), it provides better estimates than car::Anova().
type = "add1" uses stats::add1() to calculate the reduction of deviance
associated with each predictor in an univariable model (a model with just
this predictor compared to the null model). In this scenario, partial and
relative contribution are not defined. It provides an estimate of the
contribution of a variable in the absence of all other predictors.
For survey::svyglm() object, only type II and III are supported. The
type = "drop1" is recommended, in the absence of interactions, for
such models.
In a type-II or type-III Anova, the sum of relative contributions would equal 100% only if all predictors are perfectly independent. In practice, the sum is never equal to 100% due to some correlation between predictors.
Some approaches, such as dominance analysis, have been developed for decomposing the coefficient of determination (R2) into individual contributions of predictors in regression models, accounting for correlations between predictors. They address a key challenge in relative importance analysis: how to allocate shared variance when predictors are collinear.
Dominance analysis follows a systematic, computationally intensive approach based on all-subsets regression. Models of every possible combination of predictors are generated, each pair of predictors are compared by examining their incremental contributions within matching subsets. Finally, the dominance relationship is determined based on three levels: complete dominance (variable A always contributes more), conditional dominance (higher average contribution), and general dominance (weighted average across subset sizes).
The last dimension (the average contribution of each variable) provides a metric similar to semi-partial R2, and could be expressed as a proportion of the global R2 (relative contribution). The sum of these relative contributions is equal 100%.
tbl_dominance() is a quick helper to perform a dominance analysis with
dominanceanalysis::dominanceAnalysis() and to display the average total
contribution variable in a nicely formatted gt::gt() table. The function
also displays the relative contribution (i.e. the contribution of each
variable expressed as a proportion of the R2).
To be noted, alternative implementations of dominance analysis in R
include parameters::dominance_analysis() or domir::domin().
Note
Regarding η2 indicators, more details are provided in a
dedicated vignette
of the effectsize package. This vignette also presents variations of these
indicators. Some explanations are also available in the
documentation of the GAMLj
package for Jamovi.
To be noted, GAMLj highlights an potential issue regarding the computation
of total variance in effectsize::eta_squared(partial = FALSE). The
effectsize package sum the values displayed in the Anova object instead
of performing a null model. Here, we rely on total_deviance() and therefore
estimates for η2 are not equal to those performed by
effectsize::eta_squared(partial = FALSE). To be noted,
effectsize::eta_squared(partial = TRUE) (partial η2), is
not impacted by this difference in approaches.
Examples
# Linear model m <- lm(Sepal.Length ~ Sepal.Width + Species + Petal.Length, data = iris) m |> contributions() m |> tbl_contributions() m |> tbl_dominance() m |> tbl_contributions(type = 1) # GLM m2 <- glm(Survived == "Yes" ~ ., data = titanic, family = binomial) m2 |> contributions() m2 |> tbl_contributions() m2 |> tbl_contributions(decimals = 2) m2 |> tbl_contributions(show = "Relative", notes = FALSE) m2 |> tbl_dominance() # custom column labels tbl <- m2 |> tbl_contributions() colnames(tbl$`_data`) # list of columns tbl |> gt::cols_label( Predictor = "Variable", Total = "Custom label (total)", Partial = gt::html("Custom <em>label</em> with <strong>HTML</strong>"), Relative = gt::md("Custom _label_ with **markdown**") ) # interaction terms m3 <- glm( Survived == "Yes" ~ Class * Sex + Age, data = titanic, family = binomial, contrasts = list(Class = contr.sum, Sex = contr.sum, Age = contr.sum) ) m3 |> tbl_contributions(type = "III") # Survey-weighted GLM library(survey) m4 <- svyglm( Survived == "Yes" ~ Class + Sex + Age, design = srvyr::as_survey(titanic), family = quasibinomial ) m4 |> tbl_contributions(type = "drop1") m4 |> tbl_dominance()# Linear model m <- lm(Sepal.Length ~ Sepal.Width + Species + Petal.Length, data = iris) m |> contributions() m |> tbl_contributions() m |> tbl_dominance() m |> tbl_contributions(type = 1) # GLM m2 <- glm(Survived == "Yes" ~ ., data = titanic, family = binomial) m2 |> contributions() m2 |> tbl_contributions() m2 |> tbl_contributions(decimals = 2) m2 |> tbl_contributions(show = "Relative", notes = FALSE) m2 |> tbl_dominance() # custom column labels tbl <- m2 |> tbl_contributions() colnames(tbl$`_data`) # list of columns tbl |> gt::cols_label( Predictor = "Variable", Total = "Custom label (total)", Partial = gt::html("Custom <em>label</em> with <strong>HTML</strong>"), Relative = gt::md("Custom _label_ with **markdown**") ) # interaction terms m3 <- glm( Survived == "Yes" ~ Class * Sex + Age, data = titanic, family = binomial, contrasts = list(Class = contr.sum, Sex = contr.sum, Age = contr.sum) ) m3 |> tbl_contributions(type = "III") # Survey-weighted GLM library(survey) m4 <- svyglm( Survived == "Yes" ~ Class + Sex + Age, design = srvyr::as_survey(titanic), family = quasibinomial ) m4 |> tbl_contributions(type = "drop1") m4 |> tbl_dominance()
Cut a continuous variable in quartiles
Description
Convenient function to quickly cut a numeric vector into quartiles, i.e. by
applying cut(x, breaks = fivenum(x)). Variable label is preserved by
cut_quartiles().
Usage
cut_quartiles(x, include.lowest = TRUE, ...)cut_quartiles(x, include.lowest = TRUE, ...)
Arguments
x |
a numeric vector which is to be converted to a factor by cutting. |
include.lowest |
logical, indicating if an ‘x[i]’ equal to
the lowest (or highest, for |
... |
further arguments passed to |
Examples
mtcars$mpg |> cut_quartiles() |> summary()mtcars$mpg |> cut_quartiles() |> summary()
Fit indices for dominance analysis of survey-weighted GLM
Description
Provides support of survey::svyglm() for
dominanceanalysis::dominanceAnalysis().
Usage
da.svyglm.fit(original.model, newdata = NULL, ...)da.svyglm.fit(original.model, newdata = NULL, ...)
Arguments
original.model |
Original fitted model |
newdata |
Data used in update statement |
... |
ignored |
Helpers for grouped tables generated with gtsummary
Description
A series of helpers for grouped tables generated by gtsummary::tbl_stack()
or gtsummary::tbl_regression() in case of multinomial models,
multi-components models or other grouped results.
grouped_tbl_pivot_wider() allows to display results in a a wide format,
with one set of columns per group. multinom_add_global_p_pivot_wider() is
a specific case for multinomial models, when displaying global p-values in a
wide format: it calls gtsummary::add_global_p(), followed by
grouped_tbl_pivot_wider(), and then keep only the last column with p-values
(see examples). Finally, as grouped regression tables doesn't have exactly
the same structure as ungrouped tables, functions as
gtsummary::bold_labels() do not always work properly. If the grouped table
is kept in a long format, style_grouped_tbl() could be use to improve the
output by styling variable labels, levels and/or group names.
TO BE NOTED: to style group names, style_grouped_tbl() convert the
table into a gt object with gtsummary::as_gt(). This function should
therefore be used last. If the table is intended to be exported to another
format, do not use style_grouped_tbl().
Usage
grouped_tbl_pivot_wider(x) multinom_add_global_p_pivot_wider( x, ..., p_value_header = "**Likelihood-ratio test**" ) style_grouped_tbl( x, bold_groups = TRUE, uppercase_groups = TRUE, bold_labels = FALSE, italicize_labels = TRUE, indent_labels = 4L, bold_levels = FALSE, italicize_levels = FALSE, indent_levels = 8L )grouped_tbl_pivot_wider(x) multinom_add_global_p_pivot_wider( x, ..., p_value_header = "**Likelihood-ratio test**" ) style_grouped_tbl( x, bold_groups = TRUE, uppercase_groups = TRUE, bold_labels = FALSE, italicize_labels = TRUE, indent_labels = 4L, bold_levels = FALSE, italicize_levels = FALSE, indent_levels = 8L )
Arguments
x |
A grouped table generated with |
... |
Additional arguments passed to |
p_value_header |
Header for the p-value column. |
bold_groups |
Bold group group names? |
uppercase_groups |
Convert group names to upper case? |
bold_labels |
Bold variable labels? |
italicize_labels |
Italicize variable labels? |
indent_labels |
Number of spaces to indent variable labels. |
bold_levels |
Bold levels? |
italicize_levels |
Italicize levels? |
indent_levels |
Number of spaces to indent levels. |
Value
A gtsummary or a gt table.
Examples
mod <- nnet::multinom( grade ~ stage + marker + age, data = gtsummary::trial, trace = FALSE ) tbl <- mod |> gtsummary::tbl_regression(exponentiate = TRUE) tbl tbl |> grouped_tbl_pivot_wider() tbl |> multinom_add_global_p_pivot_wider() |> gtsummary::bold_labels() tbl |> style_grouped_tbl() t1 <- gtsummary::trial |> gtsummary::tbl_summary(include = grade, by = trt) t2 <- gtsummary::trial |> gtsummary::tbl_summary(include = stage, by = trt) gtsummary::tbl_stack(list(t1, t2), group_header = c("Table 1", "Table 2")) |> style_grouped_tbl()mod <- nnet::multinom( grade ~ stage + marker + age, data = gtsummary::trial, trace = FALSE ) tbl <- mod |> gtsummary::tbl_regression(exponentiate = TRUE) tbl tbl |> grouped_tbl_pivot_wider() tbl |> multinom_add_global_p_pivot_wider() |> gtsummary::bold_labels() tbl |> style_grouped_tbl() t1 <- gtsummary::trial |> gtsummary::tbl_summary(include = grade, by = trt) t2 <- gtsummary::trial |> gtsummary::tbl_summary(include = stage, by = trt) gtsummary::tbl_stack(list(t1, t2), group_header = c("Table 1", "Table 2")) |> style_grouped_tbl()
Additional tests for gtsummary
Description
See gtsummary::tests for more details on how defining custom tests.
fisher.simulate.p() implements Fisher test with computation of p-values by
Monte Carlo simulation in larger than 2×2 tables (see
stats::fisher.test()).
svyttest_oneway() is designed to compare means between sub-groups for
survey objects. It is based on survey::svyttest() for comparing 2 means,
and on svyoneway() for comparing 3 means or more.
Usage
fisher.simulate.p(data, variable, by, ...) svyttest_oneway(data, variable, by, ...)fisher.simulate.p(data, variable, by, ...) svyttest_oneway(data, variable, by, ...)
Arguments
data |
A data set. |
variable |
Name of the variable to test. |
by |
Name of the by variable. |
... |
Unused. |
Examples
library(gtsummary) trial |> tbl_summary(include = grade, by = trt) |> add_p(test = all_categorical() ~ "fisher.simulate.p") iris |> srvyr::as_survey() |> tbl_svysummary( include = Petal.Length, by = Species ) |> add_p(test = all_continuous() ~ svyttest_oneway)library(gtsummary) trial |> tbl_summary(include = grade, by = trt) |> add_p(test = all_categorical() ~ "fisher.simulate.p") iris |> srvyr::as_survey() |> tbl_svysummary( include = Petal.Length, by = Species ) |> add_p(test = all_continuous() ~ svyttest_oneway)
Themes for gtsummary
Description
Additional themes for tables generated with gtsummary.
Usage
theme_gtsummary_prop_n( prop_stat = "{p}% ({n})", prop_digits = 1, mean_sd = FALSE, cont_digits = 1, missing_text = NULL, overall_string = NULL, set_theme = TRUE ) theme_gtsummary_fisher_simulate_p(set_theme = TRUE) theme_gtsummary_unweighted_n( n_unweighted_prefix = "", n_unweighted_suffix = " obs.", prop_digits = 1, mean_sd = FALSE, cont_digits = 1, missing_text = NULL, overall_string = NULL, set_theme = TRUE ) theme_gtsummary_bold_labels(set_theme = TRUE)theme_gtsummary_prop_n( prop_stat = "{p}% ({n})", prop_digits = 1, mean_sd = FALSE, cont_digits = 1, missing_text = NULL, overall_string = NULL, set_theme = TRUE ) theme_gtsummary_fisher_simulate_p(set_theme = TRUE) theme_gtsummary_unweighted_n( n_unweighted_prefix = "", n_unweighted_suffix = " obs.", prop_digits = 1, mean_sd = FALSE, cont_digits = 1, missing_text = NULL, overall_string = NULL, set_theme = TRUE ) theme_gtsummary_bold_labels(set_theme = TRUE)
Arguments
prop_stat |
( |
prop_digits |
(non-negative |
mean_sd |
(scalar |
cont_digits |
(non-negative |
missing_text |
( |
overall_string |
( |
set_theme |
(scalar |
n_unweighted_prefix, n_unweighted_suffix
|
( |
Details
theme_gtsummary_prop_n() displays, by default, proportions before the
number of observations (between brackets). This function cannot be used
simultaneously with gtsummary::theme_gtsummary_mean_sd(), but you can use
the mean_sd = TRUE option of theme_gtsummary_prop_n().
theme_gtsummary_prop_n() also modifies default method for
gtsummary::add_ci.tbl_summary() ("wilson" for categorical variables,
"t.test", i.e. mean confidence interval, for continuous variables if
mean_sd = TRUE, "wilcox.test", i.e. confidence interval of the
pseudomedian, for continuous variables if mean_sd = FALSE).
Finally, theme_gtsummary_prop_n() also modifies default tests for
gtsummary::add_p.tbl_summary() for continuous variables if
mean_sd = TRUE ("t.test" for comparing 2 groups, or "oneway.test" for
3 groups or more). If mean_sd = FALSE, the default tests for continuous
variables remain "wilcox.test" (2 groups) or "kruskal.test" (3 groups
or more). For categorical variables, "chisq.test.no.correct" and
"fisher.test" are used by default.
See theme_gtsummary_fisher_simulate_p() to change the default test for
categorical variables.
theme_gtsummary_fisher_simulate_p() modify the default test used for
categorical variables by Fisher test, with computation of p-values by
Monte Carlo simulation in larger than 2×2 tables.
theme_gtsummary_unweighted_n() modifies default values of tables returned
by gtsummary::tbl_svysummary() and displays the unweighted number of
observations instead of the weighted n.
theme_gtsummary_unweighted_n() also modifies default method for
gtsummary::add_ci.tbl_svysummary() ("svyprop.logit" for categorical
variables, "svymean", i.e. mean confidence interval, for continuous
variables if mean_sd = TRUE, "svymedian.mean", i.e. confidence interval
of the median, for continuous variables if mean_sd = FALSE).
Finally, theme_gtsummary_unweighted_n() also modifies default tests for
gtsummary::add_p.tbl_svysummary() for continuous variables if
mean_sd = TRUE (svyttest_oneway which calls survey::svyttest() for
comparing 2 means and svyoneway() for comparing 3 means or more).
If mean_sd = FALSE, the default tests for continuous
variables remain "svy.wilcox.test" which used a designed-based Wilcoxon
test (2 groups) or Kruskal-Wallis test (3 groups or more). For categorical
variables, "svy.chisq.test"is used by default.
theme_gtsummary_bold_labels() applies automatically
gtsummary::bold_labels() to all tables generated with gtsummary.
Examples
library(gtsummary) trial |> tbl_summary(include = c(grade, age), by = trt) |> add_p() theme_gtsummary_prop_n(mean_sd = TRUE) theme_gtsummary_fisher_simulate_p() theme_gtsummary_bold_labels() trial |> tbl_summary(include = c(grade, age), by = trt) |> add_p() data("api", package = "survey") apistrat$both[1:5] <- NA apistrat |> srvyr::as_survey(strata = stype, weights = pw) |> tbl_svysummary(include = c(stype, both), by = awards) |> add_overall() theme_gtsummary_unweighted_n() apistrat |> srvyr::as_survey(strata = stype, weights = pw) |> tbl_svysummary(include = c(stype, both), by = awards) |> add_overall() gtsummary::reset_gtsummary_theme()library(gtsummary) trial |> tbl_summary(include = c(grade, age), by = trt) |> add_p() theme_gtsummary_prop_n(mean_sd = TRUE) theme_gtsummary_fisher_simulate_p() theme_gtsummary_bold_labels() trial |> tbl_summary(include = c(grade, age), by = trt) |> add_p() data("api", package = "survey") apistrat$both[1:5] <- NA apistrat |> srvyr::as_survey(strata = stype, weights = pw) |> tbl_svysummary(include = c(stype, both), by = awards) |> add_overall() theme_gtsummary_unweighted_n() apistrat |> srvyr::as_survey(strata = stype, weights = pw) |> tbl_svysummary(include = c(stype, both), by = awards) |> add_overall() gtsummary::reset_gtsummary_theme()
Utilities for gtsummary
Description
Utilities for tables generated with gtsummary.
Usage
bold_variable_group_headers(x) italicize_variable_group_headers(x) indent_levels(x, indent = 8L) indent_labels(x, indent = 4L)bold_variable_group_headers(x) italicize_variable_group_headers(x) indent_levels(x, indent = 8L) indent_labels(x, indent = 4L)
Arguments
x |
A |
indent |
An integer indicating how many space to indent text. |
See Also
gtsummary::modify_bold(), gtsummary::modify_italic(),
gtsummary::modify_indent()
Examples
library(gtsummary) tbl <- trial |> tbl_summary( include = c(stage, grade, age, trt, response, death) ) |> add_variable_group_header( header = "Clinical situation at diagnosis", variables = c(stage, grade, age) ) |> add_variable_group_header( header = "Treatment and outcome", variables = c(trt, response, death) ) tbl tbl |> bold_variable_group_headers() |> italicize_labels() |> indent_levels(indent = 8L)library(gtsummary) tbl <- trial |> tbl_summary( include = c(stage, grade, age, trt, response, death) ) |> add_variable_group_header( header = "Clinical situation at diagnosis", variables = c(stage, grade, age) ) |> add_variable_group_header( header = "Treatment and outcome", variables = c(trt, response, death) ) tbl tbl |> bold_variable_group_headers() |> italicize_labels() |> indent_levels(indent = 8L)
Install / Update project dependencies
Description
This function uses renv::dependencies() to identify R package dependencies
in a project and then calls pak::pkg_install() to install / update these
packages. If some packages are not found, the function will install those
available and returns a message indicated packages not installed/updated.
Usage
install_dependencies(dependencies = NULL, ask = TRUE)install_dependencies(dependencies = NULL, ask = TRUE)
Arguments
dependencies |
An optional list of dependencies. If |
ask |
Whether to ask for confirmation when installing a different version of a package that is already installed. Installations that only add new packages never require confirmation. |
Value
(Invisibly) A data frame with information about the installed package(s).
Examples
## Not run: install_dependencies() ## End(Not run)## Not run: install_dependencies() ## End(Not run)
Comparison tests considering NA as values to be compared
Description
is_different() and is_equal() performs comparison tests, considering
NA values as legitimate values (see examples).
Usage
is_different(x, y) is_equal(x, y) cumdifferent(x) num_cycle(x)is_different(x, y) is_equal(x, y) cumdifferent(x) num_cycle(x)
Arguments
x, y
|
Vectors to be compared. |
Details
cum_different() allows to identify groups of continuous rows that have
the same value. num_cycle() could be used to identify sub-groups that
respect a certain condition (see examples).
is_equal(x, y) is equivalent to
(x == y & !is.na(x) & !is.na(y)) | (is.na(x) & is.na(y)), and
is_different(x, y) is equivalent to
(x != y & !is.na(x) & !is.na(y)) | xor(is.na(x), is.na(y)).
Value
A vector of the same length as x.
Examples
v <- c("a", "b", NA) is_different(v, "a") is_different(v, NA) is_equal(v, "a") is_equal(v, NA) d <- dplyr::tibble(group = c("a", "a", "b", "b", "a", "b", "c", "a")) d |> dplyr::mutate( subgroup = cumdifferent(group), sub_a = num_cycle(group == "a") )v <- c("a", "b", NA) is_different(v, "a") is_different(v, NA) is_equal(v, "a") is_equal(v, NA) d <- dplyr::tibble(group = c("a", "a", "b", "b", "a", "b", "c", "a")) d |> dplyr::mutate( subgroup = cumdifferent(group), sub_a = num_cycle(group == "a") )
Add leading zeros
Description
Add leading zeros
Usage
leading_zeros(x, left_digits = NULL, digits = 0, prefix = "", suffix = "", ...)leading_zeros(x, left_digits = NULL, digits = 0, prefix = "", suffix = "", ...)
Arguments
x |
a numeric vector |
left_digits |
number of digits before decimal point, automatically computed if not provided |
digits |
number of digits after decimal point |
prefix, suffix
|
Symbols to display before and after value |
... |
additional parameters passed to |
Value
A character vector of the same length as x.
See Also
base::formatC(), base::sprintf()
Examples
v <- c(2, 103.24, 1042.147, 12.4566, NA) leading_zeros(v) leading_zeros(v, digits = 1) leading_zeros(v, left_digits = 6, big.mark = " ") leading_zeros(c(0, 6, 12, 18), prefix = "M")v <- c(2, 103.24, 1042.147, 12.4566, NA) leading_zeros(v) leading_zeros(v, digits = 1) leading_zeros(v, left_digits = 6, big.mark = " ") leading_zeros(c(0, 6, 12, 18), prefix = "M")
Transform a data frame from long format to period format
Description
Transform a data frame from long format to period format
Usage
long_to_periods(data, id, start, stop = NULL, by = NULL)long_to_periods(data, id, start, stop = NULL, by = NULL)
Arguments
data |
A data frame, or a data frame extension (e.g. a tibble). |
id |
< |
start |
< |
stop |
< |
by |
< |
Value
A tibble.
See Also
Examples
d <- dplyr::tibble( patient = c(1, 2, 3, 3, 4, 4, 4), begin = c(0, 0, 0, 1, 0, 36, 39), end = c(50, 6, 1, 16, 36, 39, 45), covar = c("no", "no", "no", "yes", "no", "yes", "yes") ) d d |> long_to_periods(id = patient, start = begin, stop = end) d |> long_to_periods(id = patient, start = begin, stop = end, by = covar) # If stop not provided, it is deduced. # However, it considers that observation ends at the last start time. d |> long_to_periods(id = patient, start = begin)d <- dplyr::tibble( patient = c(1, 2, 3, 3, 4, 4, 4), begin = c(0, 0, 0, 1, 0, 36, 39), end = c(50, 6, 1, 16, 36, 39, 45), covar = c("no", "no", "no", "yes", "no", "yes", "yes") ) d d |> long_to_periods(id = patient, start = begin, stop = end) d |> long_to_periods(id = patient, start = begin, stop = end, by = covar) # If stop not provided, it is deduced. # However, it considers that observation ends at the last start time. d |> long_to_periods(id = patient, start = begin)
Transform a data frame from long format to a sequence object
Description
Transform a data frame from long format to a sequence object
Usage
long_to_seq( data, id, time, outcome, alphabet = "auto", labels = "auto", cnames = "auto", cpal = "auto", missing.color = "#BBBBBB", ... )long_to_seq( data, id, time, outcome, alphabet = "auto", labels = "auto", cnames = "auto", cpal = "auto", missing.color = "#BBBBBB", ... )
Arguments
data |
A data frame or a data frame extension (e.g. a tibble). |
id |
< |
time |
< |
outcome |
< |
alphabet |
Optional vector containing the alphabet (the list of all
possible states).
If |
labels |
An optional vector containing state labels used for graphics.
If |
cnames |
An optional vector containing names of the different time
points. If |
cpal |
An optional colour palette for representing the states in the
graphics. If |
missing.color |
Alternative colour for representing missing values inside the sequences. |
... |
Additional arguments passed to |
Value
An object of class stslist.
See Also
Examples
library(TraMineR) # generating a data frame in long format data("biofam") d <- biofam |> dplyr::mutate(id_ind = rownames(biofam)) |> dplyr::select(id_ind, dplyr::starts_with("a")) |> tidyr::pivot_longer( cols = dplyr::starts_with("a"), names_to = "age", names_prefix = "a", values_to = "life_state" ) |> dplyr::mutate( age = as.integer(age), life_state2 = dplyr::case_when( life_state == 0 ~ "P", life_state == 1 ~ "L", life_state == 2 ~ "M", life_state == 3 ~ "LM", life_state == 4 ~ "C", life_state == 5 ~ "LC", life_state == 6 ~ "LMC", life_state == 7 ~ "D" ) ) |> labelled::set_value_labels( life_state = c( "Parent" = 0, "Left" = 1, "Married" = 2, "Left & Married" = 3, "Child" = 4, "Left & Child" = 5, "Left & Married & Child" = 6, "Divorced" = 7 ), life_state2 = c( "Parent" = "P", "Left" = "L", "Married" = "M", "Left & Married" = "LM", "Child" = "C", "Left & Child" = "LC", "Left & Married & Child" = "LMC", "Divorced" = "D" ) ) |> dplyr::mutate( life_state3 = labelled::to_factor(life_state), life_state4 = unclass(life_state2) ) d |> long_to_seq(id = id_ind, time = age, outcome = life_state) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state2) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state3) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state4) |> head(10)library(TraMineR) # generating a data frame in long format data("biofam") d <- biofam |> dplyr::mutate(id_ind = rownames(biofam)) |> dplyr::select(id_ind, dplyr::starts_with("a")) |> tidyr::pivot_longer( cols = dplyr::starts_with("a"), names_to = "age", names_prefix = "a", values_to = "life_state" ) |> dplyr::mutate( age = as.integer(age), life_state2 = dplyr::case_when( life_state == 0 ~ "P", life_state == 1 ~ "L", life_state == 2 ~ "M", life_state == 3 ~ "LM", life_state == 4 ~ "C", life_state == 5 ~ "LC", life_state == 6 ~ "LMC", life_state == 7 ~ "D" ) ) |> labelled::set_value_labels( life_state = c( "Parent" = 0, "Left" = 1, "Married" = 2, "Left & Married" = 3, "Child" = 4, "Left & Child" = 5, "Left & Married & Child" = 6, "Divorced" = 7 ), life_state2 = c( "Parent" = "P", "Left" = "L", "Married" = "M", "Left & Married" = "LM", "Child" = "C", "Left & Child" = "LC", "Left & Married & Child" = "LMC", "Divorced" = "D" ) ) |> dplyr::mutate( life_state3 = labelled::to_factor(life_state), life_state4 = unclass(life_state2) ) d |> long_to_seq(id = id_ind, time = age, outcome = life_state) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state2) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state3) |> head(10) d |> long_to_seq(id = id_ind, time = age, outcome = life_state4) |> head(10)
Compute means, standard deviations and confidence intervals by sub-groups
Description
mean_sd() lets you quickly compute mean and standard deviation by
sub-groups. Use .conf.int = TRUE to also return confidence intervals of the
mean.
Usage
mean_sd(data, ...) ## S3 method for class 'data.frame' mean_sd( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## S3 method for class 'survey.design' mean_sd( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## Default S3 method: mean_sd( data, ..., .drop = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL )mean_sd(data, ...) ## S3 method for class 'data.frame' mean_sd( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## S3 method for class 'survey.design' mean_sd( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## Default S3 method: mean_sd( data, ..., .drop = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL )
Arguments
data |
A vector, a data frame, data frame extension (e.g. a tibble), or a survey design object. |
... |
< |
.by |
< |
.drop |
If |
.drop_na_by |
If |
.conf.int |
If |
.conf.level |
Confidence level for the returned confidence intervals. |
.options |
Additional arguments passed to |
Value
A tibble. Column "n" reports the number of valid observations
and "missing" the number of missing (NA) observations, unweighted for
survey objects.
A tibble with one row per group.
Examples
# using a vector iris$Petal.Length |> mean_sd() # one variable iris |> mean_sd(Petal.Length) iris |> mean_sd(Petal.Length, .conf.int = TRUE) iris |> mean_sd(Petal.Length, .by = Species) mtcars |> mean_sd(mpg, .by = c(cyl, gear)) # two variables iris |> mean_sd(Petal.Length, Petal.Width) iris |> mean_sd(dplyr::pick(dplyr::starts_with("Petal")), .by = Species) # missing values d <- iris d$Petal.Length[1:10] <- NA d |> mean_sd(Petal.Length) d |> mean_sd(Petal.Length, .by = Species) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(iris) ds |> mean_sd(Petal.Length, .by = Species, .conf.int = TRUE)# using a vector iris$Petal.Length |> mean_sd() # one variable iris |> mean_sd(Petal.Length) iris |> mean_sd(Petal.Length, .conf.int = TRUE) iris |> mean_sd(Petal.Length, .by = Species) mtcars |> mean_sd(mpg, .by = c(cyl, gear)) # two variables iris |> mean_sd(Petal.Length, Petal.Width) iris |> mean_sd(dplyr::pick(dplyr::starts_with("Petal")), .by = Species) # missing values d <- iris d$Petal.Length[1:10] <- NA d |> mean_sd(Petal.Length) d |> mean_sd(Petal.Length, .by = Species) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(iris) ds |> mean_sd(Petal.Length, .by = Species, .conf.int = TRUE)
Compute median, quartiles and interquartile range by sub-groups
Description
median_iqr() lets you quickly compute median, quartiles and interquartile
range by sub-groups. Use .outliers = TRUE to also return whiskers and
outliers (see ggplot2::stat_boxplot()).
Usage
median_iqr(data, ...) ## S3 method for class 'data.frame' median_iqr( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .outliers = FALSE ) ## S3 method for class 'survey.design' median_iqr( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .outliers = FALSE ) ## Default S3 method: median_iqr(data, ..., .drop = FALSE, .outliers = FALSE)median_iqr(data, ...) ## S3 method for class 'data.frame' median_iqr( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .outliers = FALSE ) ## S3 method for class 'survey.design' median_iqr( data, ..., .by = NULL, .drop = FALSE, .drop_na_by = FALSE, .outliers = FALSE ) ## Default S3 method: median_iqr(data, ..., .drop = FALSE, .outliers = FALSE)
Arguments
data |
A vector, a data frame, data frame extension (e.g. a tibble), or a survey design object. |
... |
< |
.by |
< |
.drop |
If |
.drop_na_by |
If |
.outliers |
If |
Value
A tibble. Column "n" reports the number of valid observations
and "missing" the number of missing (NA) observations, unweighted for
survey objects.
A tibble with one row per group.
Examples
# using a vector iris$Petal.Length |> median_iqr() # one variable iris |> median_iqr(Petal.Length) iris |> median_iqr(Petal.Length, .outliers = TRUE) iris |> median_iqr(Petal.Length, .by = Species) mtcars |> median_iqr(mpg, .by = c(cyl, gear)) # two variables iris |> median_iqr(Petal.Length, Petal.Width) iris |> median_iqr(dplyr::pick(dplyr::starts_with("Petal")), .by = Species) # missing values d <- iris d$Petal.Length[1:10] <- NA d |> median_iqr(Petal.Length) d |> median_iqr(Petal.Length, .by = Species) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(iris) ds |> median_iqr(Petal.Length, .by = Species, .outliers = TRUE)# using a vector iris$Petal.Length |> median_iqr() # one variable iris |> median_iqr(Petal.Length) iris |> median_iqr(Petal.Length, .outliers = TRUE) iris |> median_iqr(Petal.Length, .by = Species) mtcars |> median_iqr(mpg, .by = c(cyl, gear)) # two variables iris |> median_iqr(Petal.Length, Petal.Width) iris |> median_iqr(dplyr::pick(dplyr::starts_with("Petal")), .by = Species) # missing values d <- iris d$Petal.Length[1:10] <- NA d |> median_iqr(Petal.Length) d |> median_iqr(Petal.Length, .by = Species) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(iris) ds |> median_iqr(Petal.Length, .by = Species, .outliers = TRUE)
Plot observed vs predicted distribution of a fitted model
Description
Plot observed vs predicted distribution of a fitted model
Usage
observed_vs_theoretical(model)observed_vs_theoretical(model)
Arguments
model |
A statistical model. |
Details
Has been tested with stats::lm() and stats::glm() models. It may work
with other types of models, but without any warranty.
Value
A ggplot2 plot.
Examples
# a linear model mod <- lm(Sepal.Length ~ Sepal.Width + Species, data = iris) mod |> observed_vs_theoretical() # a logistic regression mod <- glm( as.factor(Survived) ~ Class + Sex, data = titanic, family = binomial() ) mod |> observed_vs_theoretical()# a linear model mod <- lm(Sepal.Length ~ Sepal.Width + Species, data = iris) mod |> observed_vs_theoretical() # a logistic regression mod <- glm( as.factor(Survived) ~ Class + Sex, data = titanic, family = binomial() ) mod |> observed_vs_theoretical()
Transform a data frame from period format to long format
Description
Transform a data frame from period format to long format
Usage
periods_to_long( data, start, stop, time_step = 1, time_name = "time", keep = FALSE )periods_to_long( data, start, stop, time_step = 1, time_name = "time", keep = FALSE )
Arguments
data |
A data frame, or a data frame extension (e.g. a tibble). |
start |
< |
stop |
< |
time_step |
(numeric) Desired value for the time variable. |
time_name |
(character) Name of the time variable. |
keep |
(logical) Should start and stop variable be kept in the results? |
Value
A tibble.
See Also
Examples
d <- dplyr::tibble( patient = c(1, 2, 3, 3), begin = c(0, 2, 0, 3), end = c(6, 4, 2, 8), covar = c("no", "yes", "no", "yes") ) d d |> periods_to_long(start = begin, stop = end) d |> periods_to_long(start = begin, stop = end, time_step = 5)d <- dplyr::tibble( patient = c(1, 2, 3, 3), begin = c(0, 2, 0, 3), end = c(6, 4, 2, 8), covar = c("no", "yes", "no", "yes") ) d d |> periods_to_long(start = begin, stop = end) d |> periods_to_long(start = begin, stop = end, time_step = 5)
Plot a categorical variable by sub-groups
Description
Plot one or several categorical variables by sub-groups. See proportion()
for more details on the way proportions and confidence intervals are
computed. Return a bar plot (see examples).
Usage
plot_categorical( data, outcome, na.rm = TRUE, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, ..., show_overall = TRUE, overall_label = "Overall", show_pvalues = TRUE, pvalues_test = c("fisher", "chisq"), pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, pvalues_y = ifelse(flip, 1.05, 1), show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "auto", facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, return_data = FALSE )plot_categorical( data, outcome, na.rm = TRUE, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, ..., show_overall = TRUE, overall_label = "Overall", show_pvalues = TRUE, pvalues_test = c("fisher", "chisq"), pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, pvalues_y = ifelse(flip, 1.05, 1), show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "auto", facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, return_data = FALSE )
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
outcome |
< |
na.rm |
Should |
by |
< |
drop_na_by |
Remove |
convert_continuous |
Should continuous by variables (with 5 unique
values or more) be converted to quartiles (using |
... |
Additional arguments passed to |
show_overall |
Display "Overall" column? |
overall_label |
Label for the overall column. |
show_pvalues |
Display p-values in the top-left corner? |
pvalues_test |
Test to compute p-values for data frames: |
pvalues_labeller |
Labeller function for p-values. |
pvalues_size |
Text size for p-values. |
pvalues_y |
Y position of p-values. |
show_labels |
Display proportion labels? |
labels_labeller |
Labeller function for labels. |
labels_size |
Size of labels. |
labels_color |
Color of labels. |
facet_labeller |
Labeller function for strip labels. |
flip |
Flip x and y axis? |
minimal |
Should a minimal theme be applied? (no y-axis, no grid) |
return_data |
Return computed data instead of the plot? |
Examples
titanic |> plot_categorical( Class, by = c(Age, Sex) ) titanic |> plot_categorical( Class, by = c(Age, Sex), show_overall = FALSE, flip = TRUE ) titanic |> plot_categorical( Class, by = c(Age, Sex), flip = TRUE, minimal = TRUE ) gtsummary::trial |> plot_categorical(grade, by = c(age, stage, trt)) gtsummary::trial |> plot_categorical(grade, by = c(age, stage, trt), drop_na_by = TRUE) gtsummary::trial |> plot_categorical(c(grade, stage), by = c(trt, response))titanic |> plot_categorical( Class, by = c(Age, Sex) ) titanic |> plot_categorical( Class, by = c(Age, Sex), show_overall = FALSE, flip = TRUE ) titanic |> plot_categorical( Class, by = c(Age, Sex), flip = TRUE, minimal = TRUE ) gtsummary::trial |> plot_categorical(grade, by = c(age, stage, trt)) gtsummary::trial |> plot_categorical(grade, by = c(age, stage, trt), drop_na_by = TRUE) gtsummary::trial |> plot_categorical(c(grade, stage), by = c(trt, response))
Plot a continuous variable by sub-groups
Description
Plot one or several continuous variables by sub-groups. See median_iqr()
for more details on the way statistics are
computed. Return a box plot (see examples).
Usage
plot_continuous( data, outcome, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, ..., show_overall = TRUE, overall_label = "Overall", show_pvalues = TRUE, pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE )plot_continuous( data, outcome, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, ..., show_overall = TRUE, overall_label = "Overall", show_pvalues = TRUE, pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE )
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
outcome |
< |
by |
< |
drop_na_by |
Remove |
convert_continuous |
Should continuous by variables (with 5 unique
values or more) be converted to quartiles (using |
... |
Additional arguments passed to |
show_overall |
Display "Overall" column? |
overall_label |
Label for the overall column. |
show_pvalues |
Display p-values in the top-left corner? p-values are
computed with |
pvalues_labeller |
Labeller function for p-values. |
pvalues_size |
Text size for p-values. |
facet_labeller |
Labeller function for strip labels. |
flip |
Flip x and y axis? |
minimal |
Should a minimal theme be applied? (no y-axis, no grid) |
free_scale |
Allow y axis to vary between conditions? |
return_data |
Return computed data instead of the plot? |
Examples
iris |> plot_continuous(Petal.Length, by = Species) iris |> plot_continuous( dplyr::starts_with("Petal"), by = Species, free_scale = TRUE, fill = "lightblue", outlier.color = "red" ) mtcars |> plot_continuous( mpg, by = c(cyl, gear), flip = TRUE, mapping = ggplot2::aes(fill = by) ) # works with continuous by variables mtcars |> plot_continuous( mpg, by = c(disp, drat), flip = TRUE, minimal = TRUE ) # works with survey object iris |> srvyr::as_survey() |> plot_continuous( Petal.Length, by = c(Species, Petal.Width), flip = TRUE )iris |> plot_continuous(Petal.Length, by = Species) iris |> plot_continuous( dplyr::starts_with("Petal"), by = Species, free_scale = TRUE, fill = "lightblue", outlier.color = "red" ) mtcars |> plot_continuous( mpg, by = c(cyl, gear), flip = TRUE, mapping = ggplot2::aes(fill = by) ) # works with continuous by variables mtcars |> plot_continuous( mpg, by = c(disp, drat), flip = TRUE, minimal = TRUE ) # works with survey object iris |> srvyr::as_survey() |> plot_continuous( Petal.Length, by = c(Species, Petal.Width), flip = TRUE )
Plot inertia, absolute loss and relative loss from a classification tree
Description
Plot inertia, absolute loss and relative loss from a classification tree
Usage
plot_inertia_from_tree(tree, k_max = 15) get_inertia_from_tree(tree, k_max = 15)plot_inertia_from_tree(tree, k_max = 15) get_inertia_from_tree(tree, k_max = 15)
Arguments
tree |
A dendrogram, i.e. an stats::hclust object,
an FactoMineR::HCPC object or an object that can be converted to an
stats::hclust object with |
k_max |
Maximum number of clusters to return / plot. |
Value
A ggplot2 plot or a tibble.
Examples
hc <- hclust(dist(USArrests)) get_inertia_from_tree(hc) plot_inertia_from_tree(hc)hc <- hclust(dist(USArrests)) get_inertia_from_tree(hc) plot_inertia_from_tree(hc)
Plot means by sub-groups
Description
Plot one or several means by sub-groups. See mean_sd() for more details on
the way means and confidence intervals are computed.
By default, return a point plot, but other geometries could be used
(see examples).
Usage
plot_means( data, outcome, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, geom = "point", ..., show_overall = TRUE, overall_label = "Overall", show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_pvalues = TRUE, pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, show_labels = TRUE, label_y = NULL, labels_labeller = scales::label_number(0.1), labels_size = 3.5, labels_color = "black", show_overall_line = FALSE, overall_line_type = "dashed", overall_line_color = "black", overall_line_width = 0.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE )plot_means( data, outcome, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, geom = "point", ..., show_overall = TRUE, overall_label = "Overall", show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_pvalues = TRUE, pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, show_labels = TRUE, label_y = NULL, labels_labeller = scales::label_number(0.1), labels_size = 3.5, labels_color = "black", show_overall_line = FALSE, overall_line_type = "dashed", overall_line_color = "black", overall_line_width = 0.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE )
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
outcome |
< |
by |
< |
drop_na_by |
Remove |
convert_continuous |
Should continuous by variables (with 5 unique
values or more) be converted to quartiles (using |
geom |
Geometry to use for plotting means ( |
... |
Additional arguments passed to the geom defined by |
show_overall |
Display "Overall" column? |
overall_label |
Label for the overall column. |
show_ci |
Display confidence intervals? |
conf_level |
Confidence level for the confidence intervals. |
ci_color |
Color of the error bars representing confidence intervals. |
show_pvalues |
Display p-values in the top-left corner? p-values are
computed with |
pvalues_labeller |
Labeller function for p-values. |
pvalues_size |
Text size for p-values. |
show_labels |
Display mean labels? |
label_y |
Y position of labels. If |
labels_labeller |
Labeller function for labels. |
labels_size |
Size of labels. |
labels_color |
Color of labels. |
show_overall_line |
Add an overall line? |
overall_line_type |
Line type of the overall line. |
overall_line_color |
Color of the overall line. |
overall_line_width |
Line width of the overall line. |
facet_labeller |
Labeller function for strip labels. |
flip |
Flip x and y axis? |
minimal |
Should a minimal theme be applied? (no y-axis, no grid) |
free_scale |
Allow y axis to vary between conditions? |
return_data |
Return computed data instead of the plot? |
Examples
iris |> plot_means(Petal.Length, by = Species) iris |> plot_means( dplyr::starts_with("Petal"), by = Species, geom = "bar", fill = "lightblue", show_overall_line = TRUE ) mtcars |> plot_means( mpg, by = c(cyl, gear), size = 3, colour = "plum", flip = TRUE ) # works with continuous by variables mtcars |> plot_means( mpg, by = c(disp, drat), fill = "plum", geom = "bar", flip = TRUE, minimal = TRUE ) # works with survey object iris |> srvyr::as_survey() |> plot_means( Petal.Length, by = c(Species, Petal.Width), label_y = -1, size = 3, mapping = ggplot2::aes(colour = by), flip = TRUE )iris |> plot_means(Petal.Length, by = Species) iris |> plot_means( dplyr::starts_with("Petal"), by = Species, geom = "bar", fill = "lightblue", show_overall_line = TRUE ) mtcars |> plot_means( mpg, by = c(cyl, gear), size = 3, colour = "plum", flip = TRUE ) # works with continuous by variables mtcars |> plot_means( mpg, by = c(disp, drat), fill = "plum", geom = "bar", flip = TRUE, minimal = TRUE ) # works with survey object iris |> srvyr::as_survey() |> plot_means( Petal.Length, by = c(Species, Petal.Width), label_y = -1, size = 3, mapping = ggplot2::aes(colour = by), flip = TRUE )
Plot a multiple answers question
Description
Considering a multiple answers question coded as several binary variables
(one per answer), plot the proportion of positive answers.
If combine_answers = FALSE, plot the proportion of positive answers of each
item, separately. If combine_answers = FALSE, combine the different answers
(see combine_answers()) and plot the proportion of each combination
(ggupset package required when
flip = FALSE).
See proportion() for more details on the way proportions and
confidence intervals are computed. By default, return a bar plot, but other
geometries could be used (see examples). If defined, use variable labels
(see examples).
Usage
plot_multiple_answers( data, answers = dplyr::everything(), value = NULL, by = NULL, combine_answers = FALSE, combine_sep = " | ", missing_label = " missing", none_label = "none", drop_na = FALSE, drop_na_by = FALSE, sort = c("none", "ascending", "descending", "degrees"), geom = "bar", ..., show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", flip = FALSE, return_data = FALSE ) plot_multiple_answers_dodge( data, answers = dplyr::everything(), value = NULL, by, combine_answers = FALSE, combine_sep = " | ", missing_label = " missing", none_label = "none", drop_na = FALSE, drop_na_by = FALSE, sort = c("none", "ascending", "descending", "degrees"), geom = c("bar", "point"), width = 0.75, ..., show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", flip = FALSE )plot_multiple_answers( data, answers = dplyr::everything(), value = NULL, by = NULL, combine_answers = FALSE, combine_sep = " | ", missing_label = " missing", none_label = "none", drop_na = FALSE, drop_na_by = FALSE, sort = c("none", "ascending", "descending", "degrees"), geom = "bar", ..., show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", flip = FALSE, return_data = FALSE ) plot_multiple_answers_dodge( data, answers = dplyr::everything(), value = NULL, by, combine_answers = FALSE, combine_sep = " | ", missing_label = " missing", none_label = "none", drop_na = FALSE, drop_na_by = FALSE, sort = c("none", "ascending", "descending", "degrees"), geom = c("bar", "point"), width = 0.75, ..., show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_labels = TRUE, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", flip = FALSE )
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
answers |
< |
value |
Value indicating a positive answer. By default, will use the maximum observed value and will display a message. |
by |
< |
combine_answers |
Should answers be combined? (see examples) |
combine_sep |
Character string to separate combined answers. |
missing_label |
When combining answers and
|
none_label |
When combining answers and |
drop_na |
Should any observation with a least one |
drop_na_by |
If TRUE, will remove any |
sort |
Should answers be sorted according to their proportion? They could also be sorted by degrees (number of elements) when combining answers. |
geom |
Geometry to use for plotting proportions ( |
... |
Additional arguments passed to the geom defined by |
show_ci |
Display confidence intervals? |
conf_level |
Confidence level for the confidence intervals. |
ci_color |
Color of the error bars representing confidence intervals. |
show_labels |
Display proportion labels? |
labels_labeller |
Labeller function for proportion labels. |
labels_size |
Size of proportion labels. |
labels_color |
Color of proportion labels. |
flip |
Flip x and y axis? |
return_data |
Return computed data instead of the plot? |
width |
Dodging width. |
Note
If drop_na = TRUE, any observation with at least one NA value for one
item will be dropped.
If drop_na = FALSE and combine_answers = FALSE, NA values for a
specific answer are excluded the denominator when computing
proportions. Therefore, all proportions may be computed on different
population sizes.
If drop_na = FALSE and combine_answers = TRUE, any observation with at
least one NA value will be labeled with missing_label.
Examples
d <- dplyr::tibble( q1a = sample(c("y", "n"), size = 200, replace = TRUE), q1b = sample(c("y", "n", "n", NA), size = 200, replace = TRUE), q1c = sample(c("y", "y", "n"), size = 200, replace = TRUE), q1d = sample("n", size = 200, replace = TRUE) ) d |> plot_multiple_answers(q1a:q1c) d |> labelled::set_variable_labels( q1a = "apple", q1b = "banana", q1c = "chocolate", q1d = "Dijon mustard" ) |> plot_multiple_answers( value = "y", drop_na = TRUE, sort = "desc", fill = "lightblue", flip = TRUE ) d |> plot_multiple_answers( combine_answers = TRUE, value = "y", fill = "#DDCC77", drop_na = TRUE ) d |> plot_multiple_answers( combine_answers = TRUE, value = "y", flip = TRUE, mapping = ggplot2::aes(fill = prop), show.legend = FALSE ) + ggplot2::scale_fill_distiller(palette = "Spectral") d$group <- sample(c("group A", "group B"), size = 200, replace = TRUE) d |> plot_multiple_answers( answers = q1a:q1d, by = group, combine_answers = TRUE, sort = "degrees", value = "y", fill = "grey80" ) d |> plot_multiple_answers_dodge(q1a:q1d, by = group) d |> plot_multiple_answers_dodge(q1a:q1d, by = group, flip = TRUE) d |> plot_multiple_answers_dodge(q1a:q1d, by = group, combine_answers = TRUE)d <- dplyr::tibble( q1a = sample(c("y", "n"), size = 200, replace = TRUE), q1b = sample(c("y", "n", "n", NA), size = 200, replace = TRUE), q1c = sample(c("y", "y", "n"), size = 200, replace = TRUE), q1d = sample("n", size = 200, replace = TRUE) ) d |> plot_multiple_answers(q1a:q1c) d |> labelled::set_variable_labels( q1a = "apple", q1b = "banana", q1c = "chocolate", q1d = "Dijon mustard" ) |> plot_multiple_answers( value = "y", drop_na = TRUE, sort = "desc", fill = "lightblue", flip = TRUE ) d |> plot_multiple_answers( combine_answers = TRUE, value = "y", fill = "#DDCC77", drop_na = TRUE ) d |> plot_multiple_answers( combine_answers = TRUE, value = "y", flip = TRUE, mapping = ggplot2::aes(fill = prop), show.legend = FALSE ) + ggplot2::scale_fill_distiller(palette = "Spectral") d$group <- sample(c("group A", "group B"), size = 200, replace = TRUE) d |> plot_multiple_answers( answers = q1a:q1d, by = group, combine_answers = TRUE, sort = "degrees", value = "y", fill = "grey80" ) d |> plot_multiple_answers_dodge(q1a:q1d, by = group) d |> plot_multiple_answers_dodge(q1a:q1d, by = group, flip = TRUE) d |> plot_multiple_answers_dodge(q1a:q1d, by = group, combine_answers = TRUE)
Plot proportions by sub-groups
Description
Plot one or several proportions (defined by logical conditions) by
sub-groups. See proportion() for more details on the way proportions and
confidence intervals are computed. By default, return a bar plot, but other
geometries could be used (see examples). stratified_by() is an helper
function facilitating a stratified analyses (i.e. proportions by groups
stratified according to a third variable, see examples).
dummy_proportions() is an helper to easily convert a categorical variable
into dummy variables and therefore showing the proportion of each level of
the original variable (see examples).
Usage
plot_proportions( data, condition, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, geom = "bar", ..., show_overall = TRUE, overall_label = "Overall", show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_pvalues = TRUE, pvalues_test = c("fisher", "chisq"), pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, show_labels = TRUE, label_y = NULL, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", show_overall_line = FALSE, overall_line_type = "dashed", overall_line_color = "black", overall_line_width = 0.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE ) stratified_by(condition, strata) dummy_proportions(variable)plot_proportions( data, condition, by = NULL, drop_na_by = FALSE, convert_continuous = TRUE, geom = "bar", ..., show_overall = TRUE, overall_label = "Overall", show_ci = TRUE, conf_level = 0.95, ci_color = "black", show_pvalues = TRUE, pvalues_test = c("fisher", "chisq"), pvalues_labeller = scales::label_pvalue(add_p = TRUE), pvalues_size = 3.5, show_labels = TRUE, label_y = NULL, labels_labeller = scales::label_percent(1), labels_size = 3.5, labels_color = "black", show_overall_line = FALSE, overall_line_type = "dashed", overall_line_color = "black", overall_line_width = 0.5, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), flip = FALSE, minimal = FALSE, free_scale = FALSE, return_data = FALSE ) stratified_by(condition, strata) dummy_proportions(variable)
Arguments
data |
A data frame, data frame extension (e.g. a tibble), or a survey design object. |
condition |
< |
by |
< |
drop_na_by |
Remove |
convert_continuous |
Should continuous by variables (with 5 unique
values or more) be converted to quartiles (using |
geom |
Geometry to use for plotting proportions ( |
... |
Additional arguments passed to the geom defined by |
show_overall |
Display "Overall" column? |
overall_label |
Label for the overall column. |
show_ci |
Display confidence intervals? |
conf_level |
Confidence level for the confidence intervals. |
ci_color |
Color of the error bars representing confidence intervals. |
show_pvalues |
Display p-values in the top-left corner? |
pvalues_test |
Test to compute p-values for data frames: |
pvalues_labeller |
Labeller function for p-values. |
pvalues_size |
Text size for p-values. |
show_labels |
Display proportion labels? |
label_y |
Y position of labels. If |
labels_labeller |
Labeller function for labels. |
labels_size |
Size of labels. |
labels_color |
Color of labels. |
show_overall_line |
Add an overall line? |
overall_line_type |
Line type of the overall line. |
overall_line_color |
Color of the overall line. |
overall_line_width |
Line width of the overall line. |
facet_labeller |
Labeller function for strip labels. |
flip |
Flip x and y axis? |
minimal |
Should a minimal theme be applied? (no y-axis, no grid) |
free_scale |
Allow y axis to vary between conditions? |
return_data |
Return computed data instead of the plot? |
strata |
Stratification variable |
variable |
Variable to be converted into dummy variables. |
Examples
titanic |> plot_proportions( Survived == "Yes", overall_label = "All", labels_color = "white" ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue" ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue", flip = TRUE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue", minimal = TRUE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "point", color = "red", size = 3, show_labels = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "area", fill = "lightgreen", show_overall = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "line", color = "purple", ci_color = "darkblue", show_overall = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = -Survived, mapping = ggplot2::aes(fill = by), color = "black", show.legend = FALSE, show_overall_line = TRUE, show_pvalues = FALSE ) # defining several proportions titanic |> plot_proportions( dplyr::tibble( Survived = Survived == "Yes", Male = Sex == "Male" ), by = c(Class), mapping = ggplot2::aes(fill = condition) ) titanic |> plot_proportions( dplyr::tibble( Survived = Survived == "Yes", Male = Sex == "Male" ), by = c(Class), mapping = ggplot2::aes(fill = condition), free_scale = TRUE ) iris |> plot_proportions( dplyr::tibble( "Long sepal" = Sepal.Length > 6, "Short petal" = Petal.Width < 1 ), by = Species, fill = "palegreen" ) iris |> plot_proportions( dplyr::tibble( "Long sepal" = Sepal.Length > 6, "Short petal" = Petal.Width < 1 ), by = Species, fill = "palegreen", flip = TRUE ) # works with continuous by variables iris |> labelled::set_variable_labels( Sepal.Length = "Length of the sepal" ) |> plot_proportions( Species == "versicolor", by = dplyr::contains("leng"), fill = "plum", colour = "plum4" ) # works with survey object titanic |> srvyr::as_survey() |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "darksalmon", color = "black", show_overall_line = TRUE, labels_labeller = scales::label_percent(.1) ) # stratified analysis titanic |> plot_proportions( (Survived == "Yes") |> stratified_by(Sex), by = Class, mapping = ggplot2::aes(fill = condition) ) + ggplot2::theme(legend.position = "bottom") + ggplot2::labs(fill = NULL) # Convert Class into dummy variables titanic |> plot_proportions( dummy_proportions(Class), by = Sex, mapping = ggplot2::aes(fill = level) )titanic |> plot_proportions( Survived == "Yes", overall_label = "All", labels_color = "white" ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue" ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue", flip = TRUE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "lightblue", minimal = TRUE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "point", color = "red", size = 3, show_labels = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "area", fill = "lightgreen", show_overall = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = c(Class, Sex), geom = "line", color = "purple", ci_color = "darkblue", show_overall = FALSE ) titanic |> plot_proportions( Survived == "Yes", by = -Survived, mapping = ggplot2::aes(fill = by), color = "black", show.legend = FALSE, show_overall_line = TRUE, show_pvalues = FALSE ) # defining several proportions titanic |> plot_proportions( dplyr::tibble( Survived = Survived == "Yes", Male = Sex == "Male" ), by = c(Class), mapping = ggplot2::aes(fill = condition) ) titanic |> plot_proportions( dplyr::tibble( Survived = Survived == "Yes", Male = Sex == "Male" ), by = c(Class), mapping = ggplot2::aes(fill = condition), free_scale = TRUE ) iris |> plot_proportions( dplyr::tibble( "Long sepal" = Sepal.Length > 6, "Short petal" = Petal.Width < 1 ), by = Species, fill = "palegreen" ) iris |> plot_proportions( dplyr::tibble( "Long sepal" = Sepal.Length > 6, "Short petal" = Petal.Width < 1 ), by = Species, fill = "palegreen", flip = TRUE ) # works with continuous by variables iris |> labelled::set_variable_labels( Sepal.Length = "Length of the sepal" ) |> plot_proportions( Species == "versicolor", by = dplyr::contains("leng"), fill = "plum", colour = "plum4" ) # works with survey object titanic |> srvyr::as_survey() |> plot_proportions( Survived == "Yes", by = c(Class, Sex), fill = "darksalmon", color = "black", show_overall_line = TRUE, labels_labeller = scales::label_percent(.1) ) # stratified analysis titanic |> plot_proportions( (Survived == "Yes") |> stratified_by(Sex), by = Class, mapping = ggplot2::aes(fill = condition) ) + ggplot2::theme(legend.position = "bottom") + ggplot2::labs(fill = NULL) # Convert Class into dummy variables titanic |> plot_proportions( dummy_proportions(Class), by = Sex, mapping = ggplot2::aes(fill = level) )
Plot trajectories
Description
Create a trajectory index plot (similar to sequence index plot) from a data frame in long or period format.
Usage
plot_trajectories( data, id, time, fill, by = NULL, sort_by = NULL, nudge_x = NULL, hide_y_labels = NULL, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), ... ) plot_periods( data, id, start, stop, fill, by = NULL, sort_by = NULL, nudge_x = NULL, hide_y_labels = NULL, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), ... )plot_trajectories( data, id, time, fill, by = NULL, sort_by = NULL, nudge_x = NULL, hide_y_labels = NULL, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), ... ) plot_periods( data, id, start, stop, fill, by = NULL, sort_by = NULL, nudge_x = NULL, hide_y_labels = NULL, facet_labeller = ggplot2::label_wrap_gen(width = 50, multi_line = TRUE), ... )
Arguments
data |
A data frame, a data frame extension (e.g. a tibble), or a survey design object. |
id |
< |
time |
< |
fill |
< |
by |
< |
sort_by |
< |
nudge_x |
Optional amount of horizontal distance to move. |
hide_y_labels |
Hide y labels? If |
facet_labeller |
Labeller function for strip labels. |
... |
Additional arguments passed to |
start, stop
|
< |
Note
plot_trajectories() assumes that data are stored in a long format (i.e.
one row per unit of time). You can use tidyr::pivot_longer() or
periods_to_long() to transform your data in such format. By default, tiles
are centered on the value of time. You can adjust horizontal position with
nudge_x. By default, each row is assumed to represent one unit of time and
represented with a width of 1. You can adjust tiles' width with width.
plot_periods() is adapted for period format with a start and a stop
variable. You can use long_to_periods() to transform your data in such
format. Beginning and ending of each tile is determined by start and
stop arguments.
For survey design objects, weights are not taken into account. Each individual trajectory as the same height.
Examples
d <- dplyr::tibble( id = c(1, 1, 1, 1, 2, 2, 2, 3, 3, 3, 3, 3), time = c(0:3, 0:2, 0:4), status = c("a", "a", "b", "b", "b", "b", "a", "b", "b", "b", "b", "a"), group = c("f", "f", "f", "f", "f", "f", "f", "m", "m", "m", "m", "m") ) d |> plot_trajectories(id = id, time = time, fill = status, colour = "black") d |> plot_trajectories(id = id, time = time, fill = status, nudge_x = .5) d |> plot_trajectories(id = id, time = time, fill = status, by = group) d2 <- d |> dplyr::mutate(end = time + 1) |> long_to_periods(id = id, start = time, stop = end, by = status) d2 d2 |> plot_periods( id = id, start = time, stop = end, fill = status, colour = "black", height = 0.8 )d <- dplyr::tibble( id = c(1, 1, 1, 1, 2, 2, 2, 3, 3, 3, 3, 3), time = c(0:3, 0:2, 0:4), status = c("a", "a", "b", "b", "b", "b", "a", "b", "b", "b", "b", "a"), group = c("f", "f", "f", "f", "f", "f", "f", "m", "m", "m", "m", "m") ) d |> plot_trajectories(id = id, time = time, fill = status, colour = "black") d |> plot_trajectories(id = id, time = time, fill = status, nudge_x = .5) d |> plot_trajectories(id = id, time = time, fill = status, by = group) d2 <- d |> dplyr::mutate(end = time + 1) |> long_to_periods(id = id, start = time, stop = end, by = status) d2 d2 |> plot_periods( id = id, start = time, stop = end, fill = status, colour = "black", height = 0.8 )
Compute proportions
Description
proportion() lets you quickly count observations (like dplyr::count())
and compute relative proportions. Proportions are computed separately by
group (see examples).
Usage
proportion(data, ...) ## S3 method for class 'data.frame' proportion( data, ..., .by = NULL, .na.rm = FALSE, .weight = NULL, .scale = 100, .sort = FALSE, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = list(correct = TRUE) ) ## S3 method for class 'survey.design' proportion( data, ..., .by = NULL, .na.rm = FALSE, .scale = 100, .sort = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## Default S3 method: proportion( data, ..., .na.rm = FALSE, .scale = 100, .sort = FALSE, .drop = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = list(correct = TRUE) )proportion(data, ...) ## S3 method for class 'data.frame' proportion( data, ..., .by = NULL, .na.rm = FALSE, .weight = NULL, .scale = 100, .sort = FALSE, .drop = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = list(correct = TRUE) ) ## S3 method for class 'survey.design' proportion( data, ..., .by = NULL, .na.rm = FALSE, .scale = 100, .sort = FALSE, .drop_na_by = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = NULL ) ## Default S3 method: proportion( data, ..., .na.rm = FALSE, .scale = 100, .sort = FALSE, .drop = FALSE, .conf.int = FALSE, .conf.level = 0.95, .options = list(correct = TRUE) )
Arguments
data |
A vector, a data frame, data frame extension (e.g. a tibble), or a survey design object. |
... |
< |
.by |
< |
.na.rm |
Should |
.weight |
< |
.scale |
A scaling factor applied to proportion. Use |
.sort |
If |
.drop |
If |
.drop_na_by |
If |
.conf.int |
If |
.conf.level |
Confidence level for the returned confidence intervals. |
.options |
Additional arguments passed to |
Value
A tibble.
A tibble with one row per group.
Examples
# using a vector titanic$Class |> proportion() # univariable table titanic |> proportion(Class) titanic |> proportion(Class, .sort = TRUE) titanic |> proportion(Class, .conf.int = TRUE) titanic |> proportion(Class, .conf.int = TRUE, .scale = 1) # bivariable table titanic |> proportion(Class, Survived) # proportions of the total titanic |> proportion(Survived, .by = Class) # row proportions titanic |> # equivalent syntax dplyr::group_by(Class) |> proportion(Survived) # combining 3 variables or more titanic |> proportion(Class, Sex, Survived) titanic |> proportion(Sex, Survived, .by = Class) titanic |> proportion(Survived, .by = c(Class, Sex)) # missing values dna <- titanic dna$Survived[c(1:20, 500:530)] <- NA dna |> proportion(Survived) dna |> proportion(Survived, .na.rm = TRUE) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(titanic) # univariable table ds |> proportion(Class) ds |> proportion(Class, .sort = TRUE) ds |> proportion(Class, .conf.int = TRUE) ds |> proportion(Class, .conf.int = TRUE, .scale = 1) # bivariable table ds |> proportion(Class, Survived) # proportions of the total ds |> proportion(Survived, .by = Class) # row proportions ds |> dplyr::group_by(Class) |> proportion(Survived) # combining 3 variables or more ds |> proportion(Class, Sex, Survived) ds |> proportion(Sex, Survived, .by = Class) ds |> proportion(Survived, .by = c(Class, Sex)) # missing values dsna <- srvyr::as_survey(dna) dsna |> proportion(Survived) dsna |> proportion(Survived, .na.rm = TRUE)# using a vector titanic$Class |> proportion() # univariable table titanic |> proportion(Class) titanic |> proportion(Class, .sort = TRUE) titanic |> proportion(Class, .conf.int = TRUE) titanic |> proportion(Class, .conf.int = TRUE, .scale = 1) # bivariable table titanic |> proportion(Class, Survived) # proportions of the total titanic |> proportion(Survived, .by = Class) # row proportions titanic |> # equivalent syntax dplyr::group_by(Class) |> proportion(Survived) # combining 3 variables or more titanic |> proportion(Class, Sex, Survived) titanic |> proportion(Sex, Survived, .by = Class) titanic |> proportion(Survived, .by = c(Class, Sex)) # missing values dna <- titanic dna$Survived[c(1:20, 500:530)] <- NA dna |> proportion(Survived) dna |> proportion(Survived, .na.rm = TRUE) ## SURVEY DATA ------------------------------------------------------ ds <- srvyr::as_survey(titanic) # univariable table ds |> proportion(Class) ds |> proportion(Class, .sort = TRUE) ds |> proportion(Class, .conf.int = TRUE) ds |> proportion(Class, .conf.int = TRUE, .scale = 1) # bivariable table ds |> proportion(Class, Survived) # proportions of the total ds |> proportion(Survived, .by = Class) # row proportions ds |> dplyr::group_by(Class) |> proportion(Survived) # combining 3 variables or more ds |> proportion(Class, Sex, Survived) ds |> proportion(Sex, Survived, .by = Class) ds |> proportion(Survived, .by = c(Class, Sex)) # missing values dsna <- srvyr::as_survey(dna) dsna |> proportion(Survived) dsna |> proportion(Survived, .na.rm = TRUE)
Round values while preserve their rounded sum in R
Description
Sometimes, the sum of rounded numbers (e.g., using base::round()) is not
the same as their rounded sum.
Usage
round_preserve_sum(x, digits = 0)round_preserve_sum(x, digits = 0)
Arguments
x |
Numerical vector to sum. |
digits |
Number of decimals for rounding. |
Details
This solution applies the following algorithm
-
Round down to the specified number of decimal places
-
Order numbers by their remainder values
-
Increment the specified decimal place of values with k largest remainders, where k is the number of values that must be incremented to preserve their rounded sum
Value
A numerical vector of same length as x.
Source
https://biostatmatt.com/archives/2902
Examples
sum(c(0.333, 0.333, 0.334)) round(c(0.333, 0.333, 0.334), 2) sum(round(c(0.333, 0.333, 0.334), 2)) round_preserve_sum(c(0.333, 0.333, 0.334), 2) sum(round_preserve_sum(c(0.333, 0.333, 0.334), 2))sum(c(0.333, 0.333, 0.334)) round(c(0.333, 0.333, 0.334), 2) sum(round(c(0.333, 0.333, 0.334), 2)) round_preserve_sum(c(0.333, 0.333, 0.334), 2) sum(round_preserve_sum(c(0.333, 0.333, 0.334), 2))
A safe discrete colour palette
Description
Provides a safe colour palette for categorical variable. It is based on
Paul Tol's colour schemes designed to be distinct for all people, including
colour-blind readers, distinct from black and white, distinct on screen and
paper, and matching well together. It is primarily based on the bright
colour scheme implemented in khroma::scale_fill_bright(). This colour
scheme include 7 colours, including a grey reserved for NA values.
Therefore, scale_fill_safe() use the bright scheme only if 6 or less
colours are needed (keeping the grey for any NA value). If 7 to 9 colours
are needed, the muted scheme (cf. khroma::scale_fill_muted()) is used
instead. Finally, if 10 or more colours are requested, the rainbow scheme
is used (cf. khroma::scale_fill_discreterainbow()). This is a sequential
colour scheme. Here, colour are randomly reordered to provide more contrasts
between modalities.
Usage
safe_pal(reverse = FALSE) scale_fill_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "fill", na.value = "#BBBBBB" ) scale_colour_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "colour", na.value = "#BBBBBB" ) scale_color_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "colour", na.value = "#BBBBBB" )safe_pal(reverse = FALSE) scale_fill_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "fill", na.value = "#BBBBBB" ) scale_colour_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "colour", na.value = "#BBBBBB" ) scale_color_safe( name = ggplot2::waiver(), ..., reverse = FALSE, aesthetics = "colour", na.value = "#BBBBBB" )
Arguments
reverse |
A logical scalar: should the resulting vector of colours be reversed? |
name |
The name of the scale. Used as the axis or legend title.
If |
... |
Other arguments passed on to |
aesthetics |
Character string or vector of character strings listing
the name(s) of the aesthetic(s) that this scale works with. This can be
useful, for example, to apply colour settings to the colour and fill
aesthetics at the same time, via |
na.value |
Colour to be used for |
Value
A palette function.
Examples
scales::show_col(safe_pal()(6)) scales::show_col(safe_pal(reverse = TRUE)(6)) scales::show_col(safe_pal()(9)) scales::show_col(safe_pal()(16)) ggplot2::ggplot(titanic) + ggplot2::aes(x = Age, fill = Class) + ggplot2::geom_bar() + scale_fill_safe() ggplot2::ggplot(iris) + ggplot2::aes(x = Petal.Length, y = Petal.Width, colour = Species) + ggplot2::geom_point(size = 3) + scale_colour_safe()scales::show_col(safe_pal()(6)) scales::show_col(safe_pal(reverse = TRUE)(6)) scales::show_col(safe_pal()(9)) scales::show_col(safe_pal()(16)) ggplot2::ggplot(titanic) + ggplot2::aes(x = Age, fill = Class) + ggplot2::geom_bar() + scale_fill_safe() ggplot2::ggplot(iris) + ggplot2::aes(x = Petal.Length, y = Petal.Width, colour = Species) + ggplot2::geom_point(size = 3) + scale_colour_safe()
Apply step(), taking into account missing values
Description
When your data contains missing values, concerned observations are removed from a model. However, then at a later stage, you try to apply a descending stepwise approach to reduce your model by minimization of AIC, you may encounter an error because the number of rows has changed.
Usage
step_with_na(model, ...) ## Default S3 method: step_with_na(model, ..., full_data = eval(model$call$data)) ## S3 method for class 'svyglm' step_with_na(model, ..., design)step_with_na(model, ...) ## Default S3 method: step_with_na(model, ..., full_data = eval(model$call$data)) ## S3 method for class 'svyglm' step_with_na(model, ..., design)
Arguments
model |
A model object. |
... |
Additional parameters passed to |
full_data |
Full data frame used for the model, including missing data. |
design |
Survey design previously passed to |
Details
step_with_na() applies the following strategy:
-
recomputes the models using only complete cases;
-
applies
stats::step(); -
recomputes the reduced model using the full original dataset.
step_with_na() has been tested with stats::lm(), stats::glm(),
nnet::multinom(), survey::svyglm() and survival::coxph().
It may be working with other types of models, but with no warranty.
In some cases, it may be necessary to provide the full dataset initially used to estimate the model.
step_with_na() may not work inside other functions. In that case, you
may try to pass full_data to the function.
Value
The stepwise-selected model.
Examples
set.seed(42) d <- titanic |> dplyr::mutate( Group = sample( c("a", "b", NA), dplyr::n(), replace = TRUE ) ) mod <- glm(as.factor(Survived) ~ ., data = d, family = binomial()) # step(mod) should produce an error mod2 <- step_with_na(mod, full_data = d) mod2 ## WITH SURVEY --------------------------------------- library(survey) ds <- d |> dplyr::mutate(Survived = as.factor(Survived)) |> srvyr::as_survey() mods <- survey::svyglm( Survived ~ Class + Group + Sex, design = ds, family = quasibinomial() ) mod2s <- step_with_na(mods, design = ds) mod2sset.seed(42) d <- titanic |> dplyr::mutate( Group = sample( c("a", "b", NA), dplyr::n(), replace = TRUE ) ) mod <- glm(as.factor(Survived) ~ ., data = d, family = binomial()) # step(mod) should produce an error mod2 <- step_with_na(mod, full_data = d) mod2 ## WITH SURVEY --------------------------------------- library(survey) ds <- d |> dplyr::mutate(Survived = as.factor(Survived)) |> srvyr::as_survey() mods <- survey::svyglm( Survived ~ Class + Group + Sex, design = ds, family = quasibinomial() ) mod2s <- step_with_na(mods, design = ds) mod2s
Test for Equal Means for survey design object
Description
This function allows to compare several means using survey::svyglm(). More
precisely, this is a wrapper for survey::regTermTest(m, "group") where
m <- survey::svyglm(x ~ group, design).
Usage
svyoneway(formula, design, ...)svyoneway(formula, design, ...)
Arguments
formula |
a formula of the form |
design |
a survey design object |
... |
additional parameters passed to |
Value
an object of class "htest"
See Also
stats::oneway.test() for classic data frames
Examples
svyoneway( Petal.Length ~ Species, design = srvyr::as_survey(iris) )svyoneway( Petal.Length ~ Species, design = srvyr::as_survey(iris) )
Titanic data set in long format
Description
This titanic dataset is equivalent to
datasets::Titanic |> dplyr::as_tibble() |> tidyr::uncount(n).
Usage
titanictitanic
See Also
Remove row-wise grouping
Description
Remove row-wise grouping created with dplyr::rowwise() while preserving
any other grouping declared with dplyr::group_by().
Usage
unrowwise(data)unrowwise(data)
Arguments
data |
A tibble. |
Value
A tibble.
Examples
titanic |> dplyr::rowwise() titanic |> dplyr::rowwise() |> unrowwise() titanic |> dplyr::group_by(Sex, Class) |> dplyr::rowwise() titanic |> dplyr::group_by(Sex, Class) |> dplyr::rowwise() |> unrowwise()titanic |> dplyr::rowwise() titanic |> dplyr::rowwise() |> unrowwise() titanic |> dplyr::group_by(Sex, Class) |> dplyr::rowwise() titanic |> dplyr::group_by(Sex, Class) |> dplyr::rowwise() |> unrowwise()
Display the variable dictionary of a data frame in the RStudio viewer
Description
Generates an interactive variable dictionary based on labelled::look_for().
Accepts data frames, tibbles, and also survey objects.
Usage
view_dictionary(data = NULL, details = c("basic", "none", "full")) view_detailed_dictionary(data = NULL) to_DT( x, caption = NULL, column_labels = list(pos = "#", variable = "Variable", col_type = "Type", label = "Variable label", values = "Values", missing = "Missing values", unique_values = "Unique values", na_values = "User-defined missings (values)", na_range = "User-defined missings (range)") )view_dictionary(data = NULL, details = c("basic", "none", "full")) view_detailed_dictionary(data = NULL) to_DT( x, caption = NULL, column_labels = list(pos = "#", variable = "Variable", col_type = "Type", label = "Variable label", values = "Values", missing = "Missing values", unique_values = "Unique values", na_values = "User-defined missings (values)", na_range = "User-defined missings (range)") )
Arguments
data |
a data frame, a tibble or a survey object
(if |
details |
add details about each variable (see |
x |
a tibble returned by |
caption |
an optional caption for the table |
column_labels |
Optional column labels |
Details
view_dictionary() calls labelled::look_for() and applies to_DT() to
the result to produce an HTML version of the variable dictionary. If you are
using RStudio, it will be displayed by default in the Viewer pane,
allowing to have the dictionary close to your code.
view_detailed_dictionary() is similar to view_dictionary() with the
option details = "full".
These two functions are also available through dedicated addins in RStudio. To use them, select the name of a data frame, then choose View variable dictionary in the Addins menu.
Note
to_DT() is an utility to convert the result of labelled::look_for() into
a DT::datatable().
Examples
iris |> view_dictionary() iris |> labelled::look_for(details = TRUE) |> to_DT()iris |> view_dictionary() iris |> labelled::look_for(details = TRUE) |> to_DT()